Ind. Jn. of Agri. Econ. Vol.68, No.1, Jan.-March 2013 Did Crop Insurance Programmes Change the Systematic Yield Risk? Saleem Shaik* I INTRODUCTION Modeling crop yield, revenue, or loss cost ratio distributions of a farm, tehsil, district, region, or state involves estimation and identification of systematic and random component(s) of crop yields. What is systematic component of crop yield distribution? The systematic component is a known source of variation and commonly identified with high-yielding, genetically modified, disease-insect resistant varieties or other technological innovations, and changes in crop insurance programmes. The systematic component of crop yields is modeled as linear, quadratic or cubic time trend variables. 1 What is random component of crop yield? The deviation of crop yields from the estimated systematic component is random hence the random component of crop yields. Here, the focus is on the systematic component of crop yields, specifically the average systematic component (ASC) of yield, spatial systematic risk (), i.e., variation of crop yields across districts within a state in a year and temporal systematic risk (), i.e., variation of crop yield over time for each district within a state. Identification of ASC, and is important, especially when dealing with implications of crop insurance policies and its changes. For example, are technology innovations and/or crop insurance programmes that are catering to major crop growing states in India driving differential changes in ASC, and component of yields across states? Agriculture policies including crop insurance programmes are not supposed to alter acreage allocations, production or distributions. However, given the core objective of the crop insurance programme, it is anticipated to reduce production variability of the producers during times of low yield due to natural disasters. With the turn of the century and specifically since the dawn of Independence in 1947, crop insurance is increasingly viewed as an important risk management tool and the first government-supported crop insurance programme was introduced in 1972. This led to the introduction of three more major crop insurance programmes since 1972. Given this history of crop insurance programmes in India, the study attempts to estimate and examine the changes in the ASC, and of rice and wheat yields with the *Assistant Professor, Department of Agribusiness and Applied Economics, North Dakota, State University, Fargo, ND.
90 INDIAN JOURNAL OF AGRICULTURAL ECONOMICS introduction of new crop insurance programmes in 1972, 1979 and 1999. Second, did the changes in the ASC, and of rice and wheat yields suggest bias toward major crop growing states and also over time? Past studies in modeling crop yield distributions to estimate systematic component have used time-series (TS) statistical procedures that accounts for temporal variation. However, TS statistical procedure is convoluted by spatial yield correlation or spatial random variation of yields due to periodic area-wide events and changes in technology. Recently Shaik (2007 and 2010) examined the importance of accounting for spatial variation using one- and two-way panel random effects (PRE) statistical procedures on normality of crop yields. The application to Indian district level data from 1956-2002 for 15 crops and 14 states indicated accounting for spatial variation seems to change the distribution of crop yield residuals between normal and non-normal in 20 per cent, 14 per cent and 17 per cent of districts based on skewness, kurtosis and omnibus tests, respectively. Still both the TS and PRE statistical procedures encounter the limitation of accounting for the spatial correlation or spatial random variation without taking into account the hierarchical structure of districts nested in a state and country. This spatial correlation or spatial random variation observed across districts within a state or across states within a country is nested or hierarchically structured. Hence, to accurately estimate ASC, and components of crop yields, the nested hierarchically structure that accounts for similar observable/unobservable factors within each nest needs to be used. Statistical procedure like hierarchical linear models (HLM) allows examination of the importance of accounting for the nested structure of the data as well as spatial random variation on the systematic risk of crop yields. In this paper, changes in the ASC, and of rice and wheat yields is examined using district level data from 16 rice and wheat growing states in India. Section II provides a brief description of the evolution of crop insurance programmes in India and the respectively time periods. HLM statistical procedure to estimate the systematic component is presented in the third section. Section IV gives a brief description of the data. Section V deals with an empirical application to rice and wheat growing district data from India over the period 1956-2006. Finally, the study concludes with future research issues. II EVOLUTION OF INSURANCE PROGRAMMES IN INDIA According to Rejda (1995), for any insurance (including crop insurance) programme to be viable the following ideal conditions or principles of insurance (see Figure 1) should be ensured: (1) Large number of roughly homogeneous, independently insured units.
DID CROP INSURANCE PROGRAMMES CHANGE THE SYSTEMATIC YIELD RISK? 91 (2) Economically feasible premium. (3) Determinable and measurable loss. (4) Accidental and unintentional losses. (5) Calculable chance of loss. Large number of roughly homogeneous, independently insured units Economically Feasible premium CROP INSURANCE PRINCIPLES Accidental and unintentional losses Determinable and measurable loss Calculable chance of loss Figure 1. Principles of Crop Insurance Due to the covariate risk (violates the law of large numbers) faced by agriculture around the world and not only in India, it is hard for any crop insurance programme to comply with the principles or conditions of insurance. An additional issue faced by Indian agriculture is the lack of publically available historical farm or unit level production, acreage and loss data to estimate the economically feasible premium rate that matches the risks faced by the farmer on his farm and premium provided by the government. Further, based on the aggregate indemnity, premium, subsidy and liability of an insurance programme there is lack of actuarially sound premium rates to reflect the asymmetric issues like adverse selection and moral hazard, farm size, production systems, i.e., commodities grown and production practices. 2 Special attention should be paid to the adverse selection issues especially in India as farmers with consistent losses every year end up remaining in the pool by pushing out less risky farmers. The crop insurance policies are also plagued by moral hazard issues, but will not be addressed. All these issues can be addressed with publically available crop insurance data. 3 Even though faced with issues related to principles of insurance, policy makers in India pursued the potential for crop insurance programmes since the time of
92 INDIAN JOURNAL OF AGRICULTURAL ECONOMICS Independence in 1947. Interest in mitigating risks faced by agricultural producers led to the introduction of the crop insurance scheme (CIS) programme based on the "individual approach" in 1972 by General Insurance Corporation of India (GIC). The programme started specifically to provide insurance for H-4 cotton in Gujarat and extended to paddy and groundnut crops. Crop insurance based on the "individual approach" ended in 1978. Between 1972 and 1978, the CIS programme covered 3,110 farmers, collected insurance premium of Rs.0.454 million and farmers received indemnity to the amount of Rs.3.788 million. This programme was phased out due the inability of the programme to meet the principles of crop insurance and additional characteristics of Indian agriculture. A Pilot Crop Insurance Scheme (PCIS) was introduced by GIC in 1979 based on the recommendations of the report by Prof. V. M. Dandekar (for details see Dandekar, 1976 and 1985). Unlike the individual approach, the PCIS was based on the "homogeneous area" approach. The PCIS programme covered food crops including cereals and millets, oilseeds, cotton, potato and gram spread across 13 states. This PCIS programme was restricted to loanee farmers only and on a voluntary basis up to 100 per cent (later increased to 150 per cent) of the crop loan. The PCIS programme risk was shared between GIC and state government in the ratio of 2:1. A 50 per cent premium subsidy provided to small/marginal farmers was equally shared by state and central government. The PCIS programme based on the "area approach" ended in 1984. Between 1979 and 1984, the PCIS programme covered 0.627 million farmers, collected insurance premium of Rs.19.695 million and farmers received indemnity to the amount of Rs.15.705 million. The Comprehensive Crop Insurance Scheme (CCIS) was introduced in 1985 and it is an expansion of PCIS. Similar to PCIS, the CCIS was based on the "homogeneous area" approach. The CCIS programme covered food crops including cereals and millets, oilseeds and pulses spread across 15 States and 2 Union Territories (5 states opted out after few years). This CCIS programme was restricted to loanee farmers only and on a voluntary basis up to 100 per cent of the crop loan or a maximum of Rs.10,000 per farmer. The CCIS programme premium and indemnities was shared between central and state government in the ratio of 2:1. A 50 per cent premium subsidy was provided to small/marginal farmers by the state and central government on 50:50 basis. The CCIS programme ended in 1999. Between 1985 and 1999, the CCIS programme covered 7.63 million farmers, collected insurance premium of Rs.40.36 million and farmers received indemnity to the amount of Rs.231.9 million. During the rabi season of 1999, the National Agricultural Insurance Scheme (NAIS) was introduced. Agricultural Insurance Company of India Ltd. (AIC) was formed in 2002 and by early 2003 was in-charge of NAIS programme in India. The NCIS is based on both area approach for widespread calamities, and individual approach for localised calamities such as hailstorm, landslide, cyclone and floods. The NCIS programme covers all food grains, oilseeds and annual
DID CROP INSURANCE PROGRAMMES CHANGE THE SYSTEMATIC YIELD RISK? 93 horticultural/commercial crops for which past yield data are available for an adequate number of years and available for all states and union territories. The unit of insurance varies and is defined by the state government. It could be village panchayat, mandal, hobli, circle, phirka, block, or taluka. The sum insured under NCIS programme covers the amount of loan for loanee farmers, and can also be extended to cover 150 per cent of the average yield for loanee and non-loanee farmers. The NCIS program indemnities beyond 100 per cent of premium collected will be between central and state government on 50:50 basis. For annual horticultural/ commercial crops, indemnities beyond 150 per cent of premium in the first 3 or 5 years will be shared by central and state government on 50:50 basis. Until 2006-07, the NCIS programme covered 790.8 million farmers and collected insurance premium of Rs.29.44 billion. The farmers received indemnity to the amount of Rs.98.57 billion. These five crop insurance program periods form the basis for comparison of systematic component of yield distribution using HLM. Next, the HLM statistical procedure to model yield distributions for the different phases of crop insurance programmes period is presented. III MODELING TO ESTIMATE SYSTEMATIC TEMPORAL AND SPATIAL YIELD RISK The section provides an overview of the HLM statistical procedure for modeling crop yields to estimate average systematic component, spatial and temporal systematic yield risk. Suppose y ik,t represents a 1 T vector of crop yields produced in district, i = 1,..., I, which is located in state, k = 1,...,K with t = 1,..., T data points, and x m ik,t represents a M T matrix of exogenous linear, quadratic and cubic time trend variables, m = 1,...,M; α is the intercept, β m is the associated parameters of linear, quadratic and cubic time trend variables to be estimated; and ε ik,t pure random error. The parameter coefficients in equation (1) are estimated. 3.1 Hierarchical Linear Model Statistical Procedure The HLM statistical procedure allows to model and account the nested or hierarchical structure of the crop yield data. The two-way hierarchical linear model (HLM2) statistical procedure requires the randomisation of district, i = 1,..., I nested within each state, k = 1,...,K and as: yik,t = α + β m x m ik,t + z i(k) + z k + ε ik,t.(1) where z i(k) represents district random error nested within state, k, z k represents state random error and α ik,t represents the remaining pure random error, and α is the intercept, β is the parameter of time trend variables representing the degree of polynomial.
94 INDIAN JOURNAL OF AGRICULTURAL ECONOMICS Equation (1) allows the decomposition of crop yields into systematic and random components and re-written as y ik,t = α + β γ x γ ik,t + z i(k) + z k + ε ik,t = systematic + random.(2) where ŷ ik,t =(α + β γ x γ ik,t + z i(k) + z k ) is the systematic component or predicted yields and the deviation away from the systematic component is the random component, εik,t = y ik,t ŷ ik,t. Next the ASC, and are computed based on the systematic component or predicted yields estimated from equation (2) for five crop insurance programme periods (CIPP) by state. The five crop insurance programme periods are CIPP1: Pre-crop insurance period from 1956 to 1971. CIPP2: Crop insurance scheme (CIS) from 1972 to 1978. CIPP3: Pilot crop insurance scheme (PCIS) from 1979 to 1984. CIPP4: Comprehensive Crop Insurance Scheme (CCIS) from 1985 to 1999. CIPP5: National Agriculture Insurance Scheme (NAIS) from 1999 to 2006. The average systematic component for each state, ASC k is defined as the average of the systematic yield component across all the districts in each state computed for each of the five crop insurance program periods over the period 1956-2006. This is represented as ASC k I T ŷik,t = i= i t= 1 I *T.(3) The and is defined as the coefficient of variation of temporally invariant spatial yield risk and spatially invariant temporal yield risk, respectively, of the systematic component, Ŷ ik,t computed for each state, k = 1,...,K. The coefficient of variation of systematic spatial risk for each state, k, defined as the ratio of the standard deviation over mean of temporally invariant spatial yield risk is presented as I σ (yik, t ŷik, t)/i 1 ŷik = = i = 1 k.(4) μ I ŷik ŷik, t/i i = 1 Similarly, the coefficient of variation of systematic temporal risk for each state, k, defined as the ratio of the standard deviation over mean of spatially invariant temporal yield risk is presented as
DID CROP INSURANCE PROGRAMMES CHANGE THE SYSTEMATIC YIELD RISK? 95 T σ (yk,t ŷk, t)/t 1 ŷt STR = = t = 1 k.(5) μ T ŷt ŷk, t/t t = 1 The average, temporal and spatial systematic risk is computed for the five CIPP s. IV INDIAN DISTRICT DATA The district is the smallest administrative unit in India for which data on crops are available. This study covers a total of 265 districts across rice and wheat and 16 states in India for the period 1956-2006. This is a unique data set as it provides historical district level yield that has never been used in the estimation. District level data is available from four publications - Area and Production of Principal Crops in India; Agricultural Situation in India; Statistical Abstracts of India; and Crop and Season Reports by individual States. V INDIAN STATE LEVEL ANALYSIS OF SYSTEMATIC YIELD RISKS To evaluate the changes in the average systematic component, spatial systematic risk and temporal systematic risk components, the following steps were involved. (1) Estimated and decomposed crop yields into systematic and random components (Equation 2). (2) Systematic component is used to compute average systematic component or ASC (Equation 3); spatial systematic risk or (Equation 4); and temporal systematic risk or (Equation 5). (3) The ASC, and measures are estimated for the five-crop insurance programme periods (CIPP) by state, see Figure 2. The HLM statistical procedure is used to estimate the systematic component of rice and wheat yields (Equation 2) used in the computation of ASC, and measures. The averages and coefficient of variation by five crop insurance programme periods (CIPP) for 16 rice and wheat growing states are presented in Tables 1 to 4. The results indicated the average systematic yield component increased from thepre-crop insurance period, 1956-1971 to the current NAIS programme period, 1999-2006 for rice and wheat in all the states. The only exception was Uttaranchal for rice and wheat crops, with lower average systematic yield component during the PCIS programme period, 1979 to 1984.
96 INDIAN JOURNAL OF AGRICULTURAL ECONOMICS TABLE 1. STATE-WISE AVERAGE AND CV OF SYSTEMATIC YIELD RISK BY CROP INSURANCE PROGRAMMES State (1) Time period (2) (3) (4) Rice Wheat (5) (6) (7) (8) (9) (10) Andhra Pradesh CIPP1 20 16 1.448 20.241 4.839 12 16 0.088 235.232 228.091 CIPP2 20 7 1.671 17.137 3.733 12 7 0.499 41.351 20.903 CIPP3 20 6 1.87 15.32 3.152 11 6 0.74 19.76 11.903 CIPP4 20 14 2.154 13.297 4.809 11 14 1.06 12.585 10.139 CIPP5 20 8 2.331 12.295 0.594 11 7 1.308 10.63 2.163 Bihar CIPP1 9 16 0.57 28.848 9.737 9 16 0.716 17.273 18.999 CIPP2 9 7 0.797 20.642 7.831 9 7 1.132 10.924 8.315 CIPP3 9 6 0.995 16.528 5.923 9 6 1.417 8.725 5.728 CIPP4 9 10 1.233 13.341 6.695 9 10 1.731 7.145 6.043 CIPP5 9 7 1.456 11.297 0.784 9 7 1.987 6.222 0.482 Chhattisgarh CIPP1 6 16 0.502 18.185 11.057 6 16 0.211 74.193 64.433 CIPP2 6 7 0.729 12.529 8.563 6 7 0.627 24.973 15,008 CIPP3 6 6 0.927 9.849 6.358 6 6 0.912 17.165 8.897 CIPP4 6 14 1.211 7.535 8.552 6 14 1.283 12.207 10.017 CIPP5 6 6 1.387 6.582 0.8888 6 5 1.488 10.529 0.355 Gujarat CIPP1 17 16 0.741 35.262 7.487 18 16 1,136 31.108 12.007 CIPP2 17 7 0.97 27.062 6.84 18 7 1.543 23.138 6.105 CIPP3 17 6 1.166 22.416 5.053 18 6 1.83 19.5 4.425 CIPP4 17 13 1.428 17.959 6.508 18 13 2.196 16.262 5.874 CIPP5 13 6 1.59 18.217 0.92 18 6 2.401 14.869 0.402 (11) (12)
DID CROP INSURANCE PROGRAMMES CHANGE THE SYSTEMATIC YIELD RISK? 97 TABLE 2. STATE-WISE AVERAGE AND CV OF SYSTEMATIC YIELD RISK BY CROP INSURANCE PROGRAMMES State (1) Time period (2) (3) (4) Rice Wheat (5) (6) (7) (8) (9) (10) Haryana CIPP1 6 16 1.572 15.267 3.069 7 16 1.871 12.639 8.585 CIPP2 6 7 1.788 13.585 3.489 7 7 2.259 9.146 4.167 CIPP3 6 6 1.987 12.229 2.966 7 6 2.544 8.121 3.191 CIPP4 6 11 2.24 10.845 4.203 7 11 2.877 7.182 4.098 CIPP5 6 6 2.446 9.93 0.503 7 6 3.116 6.63 0.299 Jharkhand CIPP1 5 16 0.57 28.302 9.739 5 16 0.562 20.059 24.211 CIPP2 5 7 0.797 20.25 7.832 5 7 0.978 11.524 9.626 CIPP3 5 6 0.995 16.214 5.923 5 6 1.264 8.991 6.6 CIPP4 5 10 1.213 10.824 5.26 5 10 1.552 5.809 5.964 CIPP5 Karnataka CIPP1 19 16 1.461 30.492 3.8 14 16 0.276 135.559 194.977 CIPP2 19 7 1.687 26.397 3.697 14 7 0.638 47.239 15.025 CIPP3 19 6 1.886 23.622 3.125 14 6 0.919 32.739 8.877 CIPP4 19 13 2.159 20.628 4.588 14 13 1.271 22.945 8.366 CIPP5 19 7 2.346 18.983 0.486 11 7 1.397 17.91 1.834 Madhya Pradesh CIPP1 37 16 0.305 89.9 16.984 37 16 0.518 60.64 26.311 CIPP2 37 7 0.53 52.035 11.763 37 7 0.93 33.738 10.121 CIPP3 37 6 0.729 37.876 8.087 37 6 1.215 25.819 6.68 CIPP4 37 14 1.008 27.651 10.558 37 14 1.586 19.786 8.102 CIPP5 37 6 1.189 23.218 1.036 37 6 1.787 17.558 0.521 (11) (12)
98 INDIAN JOURNAL OF AGRICULTURAL ECONOMICS TABLE 3. STATE-WISE AVERAGE AND CV OF SYSTEMATIC YIELD RISK BY CROP INSURANCE PROGRAMMES State (1) Time period (2) (3) (4) Rice Wheat (5) (6) (7) (8) (9) (10) Maharashra CIPP1 26 16 0.669 61.891 9.983 23 16 0.236 68.605 57.553 CIPP2 26 7 0.893 45.969 6.987 23 7 0.651 24.877 14.403 CIPP3 26 6 1.091 37.617 5.4 22 6 0.938 17.669 8.656 CIPP4 26 14 1.373 29.815 7.733 23 14 1.304 12.538 9.842 CIPP5 24 6 1.543 26.391 1.083 23 7 1.507 10.765 0.575 Orissa CIPP1 13 16 0.629 19.022 8.826 13 16 0.728 20.1 18.91 CIPP2 13 7 0.855 13.983 7.293 13 7 1.141 12.699 8.248 CIPP3 13 6 1.054 11.352 5.592 13 6 1.427 10.16 5.691 CIPP4 13 11 1.32 9.06 8.155 13 11 1.775 8.167 7.528 CIPP5 13 7 1.515 7.895 0.782 13 7 1.998 7.244 0.649 Punjab CIPP1 11 16 2.112 12.035 2.829 11 16 2.184 13.108 5.76 CIPP2 11 7 2.333 10.735 2.674 11 7 2.586 11.399 3.639 CIPP3 11 6 2.531 9.894 2.328 11 6 2.872 10.267 2.827 CIPP4 11 14 2.816 8.894 3.679 11 14 3.242 9.094 3.963 CIPP5 11 7 2.993 8.369 0.396 11 7 3.443 8.563 0.248 Rajasthan CIPP1 16 16 0.611 51.259 9.417 26 16 1 24.223 13.503 CIPP2 17 7 0.838 35.952 7.207 26 7 1.424 16.851 6.58 CIPP3 17 6 1.026 29.437 5.711 26 6 1.708 14.059 4.753 CIPP4 17 12 1.288 23.379 7.533 26 12 2.055 11.685 5.981 CIPP5 17 7 1.481 20.379 0.867 26 7 2.278 10.541 0.421 (11) (12)
DID CROP INSURANCE PROGRAMMES CHANGE THE SYSTEMATIC YIELD RISK? 99 TABLE 4. STATE-WISE AVERAGE AND CV OF SYSTEMATIC YIELD RISK BY CROP INSURANCE PROGRAMMES State (1) Time period (2) (3) (4) Rice Wheat (5) Tamil Nadu CIPP1 11 16 1.901 23.099 2.919 CIPP2 11 7 2.128 20.639 2.932 CIPP3 11 6 2.326 18.88 2.533 CIPP4 11 13 2.6 16.893 3.81 CIPP5 11 7 2.788 15.755 0.425 (6) (7) (8) (9) (10) Uttar Pradesh CIPP1 46 16 0.817 28.655 6.792 46 16 1.082 26.09 12.566 CIPP2 46 7 1.044 22.434 5.977 46 7 1.499 18.846 6.282 CIPP3 46 6 1.242 18.853 4.744 46 6 1.784 15.832 4.551 CIPP4 46 12 1.504 15.567 6.25 46 14 2.154 13.109 5.964 CIPP5 46 5 1.7 13.777 0.719 46 5 2.355 11.989 0.307 Uttaranchal CIPP1 2 16 1.254 35.098 4.427 2 16 0.889 41.527 15.305 CIPP2 2 7 1.48 29.725 4.215 2 7 1.305 28.285 7.214 CIPP3 7 6 1.212 31.481 4.862 7 6 1.088 35.112 7.46 CIPP4 7 12 1.474 25.882 6.378 7 14 1.459 26.193 8.808 CIPP5 West Bengal CIPP1 15 16 1.101 23.008 5.043 15 16 0.884 15.19 15.354 CIPP2 15 7 1.327 19.08 4.7 15 7 1.299 10.346 7.245 CIPP3 15 6 1.526 16.601 3.863 15 6 1.584 8.483 5.123 CIPP4 15 34 1.799 14.076 5.506 15 13 1.942 6.922 6.363 CIPP5 15 7 1.987 12.746 0.596 15 7 2.156 6.234 0.395 (11) (12)
100 INDIAN JOURNAL OF AGRICULTURAL ECONOMICS Crop Insurance in India Crop Insurance Scheme (CIS), 1972-1978 Based on individual approach and introduced by General Insurance Corporation of India. Program covered H-4 cotton in Gujarat and extended to paddy and groundnut crops and states. Pilot Crop Insurance Scheme (PCIS), 1979-1984 Based on homogeneous area approach and introduced by General Insurance Corporation of India. Program covered cereals, millets, oilseeds, cotton, potato and gram spread across 13 states. Program was restricted to loanee farmers. Comprehensive Crop Insurance Scheme (CCIS), 1985-1998 Based on homogeneous area approach and introduced by General Insurance Corporation of India. Program covered cereals, millets, oilseeds and pulses spread across 15 States and 2 Union Territories (5 states opted out after few years). Program was restricted to loanee farmers up to 100% of the crop loan or maximum of Rs. 10,000 per farmer. 5.1 Rice National Agricultural Insurance Scheme (NAIS), 1999 - current Based on homogeneous area approach for widespread calamities, and individual approach for localised calamities. Since 2003 the NAIS program is serviced by Agricultural Insurance Company of India Ltd (AIC). Programme covers all food grains, oilseeds and annual horticultural / commercial crops and available for all states and union territories. The unit of insurance varies and defined by the state government. Program was restricted to loanee and non-loanee farmers. Figure 2. Evolution of Crop Insurance Programmes in India The average systematic component of crop yields for Andhra Pradesh, Haryana, Karnataka, Punjab, Tamil Nadu and West Bengal were higher than all India average for all the five CIPP s. This suggests, the farmers in the above six states have realised higher crop yields per acre during each of the five-crop insurance programme periods by utilising the available technology. However, crop yields could be increasing at an increasing or decreasing rate, decreasing at a decreasing rate or must have reached a plateau. Further, at this stage there is no way of differentiating the increase or decrease to difference in farm size groups within each district due to the lack of historical data. The average systematic component is lower in Bihar, Chhattisgarh, Gujarat, Jharkhand, Madhya Pradesh, Maharashtra, Orissa, Rajasthan and Uttar Pradesh compared to all India average. It is possible that farmers in these states have declining yield per acre due to the lack of irrigated water and are faced with inclement weather. Second, it could also mean there is lack of emphasis by the government to push additional new crop insurance policies specifically catering to the needs and
DID CROP INSURANCE PROGRAMMES CHANGE THE SYSTEMATIC YIELD RISK? 101 prevailing conditions of the farmers in these states. With the five year plans and the associated policies in the plans including crop insurance programmes, the strategy of the Indian government was, is and will be to address the needs of major producing states and larger producing farmers. The only exception was Uttaranchal, with the average systematic yield component higher than all India average during the pre-crop insurance and CIS program period. What does it mean if the average systematic yield is increasing? This suggests the technological innovations during each of the crop insurance programme periods have increased the average yield per acre across all the districts in each state. Next, the results of the systematic component of yield risk decomposed into and is discussed. In all the 16 states, the of rice yield decreased starting from pre-crop insurance period to the present NAIS programme period. This is an indication that the differences in the systematic variation or risk of rice yields between districts within each state is declining as reflected by. A comparison across states reveals that Gujarat, Karnataka, Madhya Pradesh, Maharashtra, Rajasthan, and Uttaranchal had higher of rice yield compared to all India. This indicates an increased risk in rice crop yields across districts in each of the above 6 states. This could be due to low and high yield producing districts along with the presence of small, medium and large farmers in each state or farmers are facing increased risk due to natural disasters. For the remaining states, the of rice yield was lower than for all India. The only exception was Tamil Nadu, with higher average than the all India average during the pre-crop insurance, CIS and PCIS periods, while realised lower average than all India average during the CCIS and NAIS periods. Similarly, of rice yield declined from the pre-crop insurance period to the CPIS programme period in all the 16 states with an exception. During the CCIS programme period, the of rice yield actually increased compared to the earlier programme period. However, the was less than 1 per cent during the present NAIS programme period. The of rice yields in Andhra Pradesh, Haryana, Karnataka, Punjab, Tamil Nadu, Uttar Pradesh and West Bengal had lower of rice yield compared to all India. This is an indication that the average variation over time is declining due to the technology becoming more homogenous as well as adoption of technological innovations, lowering the risks faced by the farmers in these states. The remaining states had higher or mixed results over the different crop insurance programme periods. 5.2 Wheat Next for the wheat crop, the average systematic yield component was higher in Gujarat, Haryana, Punjab, Rajasthan, Uttar Pradesh and West Bengal compared to all India average. Once again this suggests the emphasis of the Indian government towards the needs of major producing states and associated larger producing farmers.
102 INDIAN JOURNAL OF AGRICULTURAL ECONOMICS It was lower in the remaining states with the exception of Uttaranchal where the average systematic yield component was lower than the all India average during the PCIS, CCIS and NAIS programme periods. During the pre-crop insurance and CIS periods, the average systematic component of yield was higher than all India average. In all the 15 states, the of wheat yield indicated a declining trend starting from pre-crop insurance period to the present NAIS programme period. This is an indication that the difference in the systematic yield risk of wheat yield between districts within each state is declining. A comparison across states reveals Karnataka and Madhya Pradesh had higher of wheat yield compared to all India. This indicates the variation in wheat crop yields across districts in these states is high. Bihar, Haryana, Jharkhand, Orissa, Punjab, Rajasthan and West Bengal, the of wheat yield was lower than all India. In the remaining states of Andhra Pradesh and Chhattisgarh the of wheat yields was higher than all India average during the pre-crop insurance, CIS and PCIS programme period, while it was lower during the CCIS and NAIS periods. In Gujarat, the of wheat yields was higher than the all India average during all the periods with the exception of pre-crop insurance programme period. In Maharashtra, the of wheat yields was lower than all India average during the CCIS programme period. In Uttaranchal, the of wheat yields was higher than all India average during the CIS, PCIS, and CCIS programme periods and lower in the remaining two periods. Similarly, the of wheat yield declined from the pre-crop insurance period to the CPIS programme period in all the 15 states with an exception during the CCIS programme period. The of wheat yield actually increased compared to the earlier programme period. The of wheat yield was less than 1 per cent during the present NAIS programme period. With respect to the of wheat yields, it was lower in Andhra Pradesh, Haryana, Karnataka, Punjab, Tamil Nadu, Uttar Pradesh and West Bengal compared to all India. Overall, the average systematic component of rice and wheat yield has increased for all the states. This suggests that crop yields per acre are at least increasing with technological changes over time. However, the spatial systematic component of rice and wheat yields per acre is declining across districts in most of the states with few exceptions. Similarly the temporal systematic component of rice and wheat yields per acre is declining over time in most of the districts of each state with few exceptions. VI SUMMARY AND CONCLUSIONS The study research examines the changes in the average systematic component of crop yields, spatial and temporal systematic component of yield variation or risk for rice and wheat. The ASC, and STR changes are estimated by five-crop insurance programme periods using district level data from 16 major rice and wheat growing states in India for the period, 1956-2006. Second, the HLM statistical procedure is
DID CROP INSURANCE PROGRAMMES CHANGE THE SYSTEMATIC YIELD RISK? 103 used in the estimation of the systematic component of crop yields as it accounts for hierarchical or nest structure of the spatial random variation across districts within state and across states within India. To summarise, first the average systematic component has increased with each crop insurance programme period over time for rice and wheat yield during the period, 1956-2006. The lowest was during the pre-crop insurance programme period and highest during the NAIS programme period. Second the spatial systematic risk was highest during the pre-crop insurance programme period followed by the CIS, PCIS, CCIS and NAIS programme periods for both rice and wheat crop yields. In contrast, the temporal systematic risk was the highest during the pre-crop insurance programme period followed by CIS and PCIS. During the CCIS, the temporal systematic risk was higher than PCIS. However, the temporal systematic risk was the lowest during the NAIS program period for both rice and wheat crop yields. A comparison across states revealed that the spatial and temporal risk was higher for lower acreage states (fringe production regions) compared to higher acreage states (core production regions). Future research needs to examine the importance of changes in the climate variables on systematic and random yield risk. Second, there needs to be a comparison of the changes in systematic and random yield risk due to changes in liability, premium and indemnity payments. Received July 2012. Revision accepted March 2013. NOTES 1. It is easy to estimate using nonlinear or spline trends using semi- or non-parametric techniques. However, it is simple, straight forward and logical to estimate linear (increasing), quadratic (increasing and then decreasing or decreasing and then increasing or cubic (did the yield reach a plateau or flat top), unless there is very strong a priori indication of truly a nonlinear or spline trend. 2. The last decade saw a push of weather index insurance policy by domestic policies with the support of the World Bank and private companies to address losses in crop yield due to short term variation or risk. In reality, insurance policies that provide protection due to externalities associated with extreme events like monsoons and drought. This can be identified with measures of kurtosis and skewness of crop yields distributions. Second, there is hardly consistent and significant correlation between climate variables (temperature and precipitation) and yield variation (Shaik, 2012 - unpublished manuscript). 3. The author had contacted Agricultural Insurance Company of India Ltd. to request historical data without much luck. However, this information is provided to World Bank and private companies that are pushing for weather index insurance without good evidence of correlation between climate variables and yield distributions. REFERENCES Dandekar, V.M. (1976),"Crop Insurance in India", Economic and Political Weekly, pp.a-61 to A-80. Dandekar, V.M. (1985),"Crop Insurance in India - A Review, 1976-77 to 1984-5", Economic and Political Weekly, Vol.20, Nos.25&26, June 22-29, pp.a-46 to A-59. Rejda, G.E. (1995), Principles of Risk Management and Insurance, Harper Collins College Publishers, New York. Shaik, S. (2007), "Examining the Normality of Indian Rice and Wheat Yields", Pravartak, Vol.3 pp.112-119. Shaik, S. (2010), "Importance of Panel Analysis in Examining Normality of Crop Yield Distributions", Journal of Quantitative Economics, Vol.8, No.1, pp. 55-68.